Peking University's SimForcing distills simulation motion priors into robot world models
PekingUniversity · hf · 2026-10-16
Peking University researchers present SimForcing, a simulation-guided framework for action-conditioned robot world models that must respond precisely to robot trajectories while keeping realistic visual dynamics.
- Problem: real robot videos teach visual dynamics and simulation offers structured motion supervision, but appearance differences hinder direct transfer, and inaccurate simulation predictions can mislead real-video generation
- First, motion knowledge is transferred from a simulation teacher via latent-motion distillation, aligning temporal changes in latent space to internalize motion priors while mitigating appearance mismatch
- Second, multi-block simulation conditioning with condition dropout uses simulation predictions as a controllable reference during generation
- The framework generates realistic action-conditioned videos across robot datasets
More from Embodied
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- ~25,000 Humanoid Robots Shipped in H1 2026, Over 95% from Chinese Vendors — ingliguori · 2026-10-16
- openpilot brings stoplight-stopping autonomy to a 2019 Corolla — k7agar · 2026-10-16
- Hironobu Fujiyoshi publishes deck: from AI agents to Agentic Robotics — HirokatuKataoka · 2026-10-16
- CoDA co-evolves simulation and robot policies via shared failure memory, lifting OOD success from 5.3% to 27.6% — PLAN-Lab · 2026-10-16
- Beff Jezos claims rocket Cybertruck will be the first Tesla/SpaceX lunar vehicle — beffjezos · 2026-10-16